Changes in groundwater levels

Data transfer of CORRECTIV.Local Excerpt from https://github.com/correctiv/grundwasser-data: Groundwater data The groundwater in Germany is falling dramatically. For the first time, CORRECTIV compiled an overview for data from around 6,700 groundwater measuring stations and evaluated them in comparison. Methodology Our analysis is based on raw data from groundwater measuring stations from 13 federal states. We have collected the data through a combination of scraping, i.e. the downloading of data from the websites of the authorities, and press inquiries. The groundwater measuring stations we evaluated are at different depths. In some federal states, groundwater levels are measured daily. In others, every second week or once a month is measured. In order to normalise the data, we calculated the average groundwater level for each month between 1990 and 2021. If less than 95 percent of the monthly data were available for a measuring station, we excluded them from our analysis. According to this criterion, the data situation in Bremen, Hamburg and the Saarland was not sufficient for our evaluation. In addition, we conducted a semi-automatic test for data accuracy. For this we used the Changepoint package in R. We examined jumping changes in the value history for each measuring site, which can indicate errors in the data. For example, if a measuring point is replaced and the measurements are not recalibrated for the new height, it may seem that the groundwater level has suddenly risen. We then visually checked the data for each measuring point where such jumps were identified and removed the faulty stations from our analysis. Finally, we calculated the 32-year trend using the Mann-Kendall trend test, which checks for long-term rising or falling trends. We used the trend-free pre-whitening variant of the test. The analysis revealed the trend as a change in meters per month. In order to normalise the data, we then divided the trend for each measuring point by the range of values of the measuring point (difference between the highest and lowest monthly average) and calculated it high to one year in order to express the trend in percent per year.